[R] R: Debian package and source

[Dirk Eddelbuettel]

On Wed, Jun 18, 2003 at 07:11:19PM +0200, Peter Dalgaard BSA wrote:
> "David Andel" <andel at ifi.unizh.ch> writes:
> 
> > Hi
> > 
> > As Dirk said before, if you're running unstable, the official
> > version is surely the best. If you're running testing though, you're
> > probably better off using the apt sources listed in
> > http://cran.us.r-project.org/bin/linux/debian/ReadMe since there are
> > some dependency hassles with installing the unstable version on
> > testing (at least I got tired of going through all the dependency
> > stuff).

Almost -- but we have not yet built the separate testing versions for CRAN,
and may not as a) Doug is very busy right now and b) we may not need to, see
below.

> > 
> 
> BTW, is there a good reason to keep the "testing" distribution at
> 1.6.0? I would assume that at least you'd want to include patches up
> to 1.6.2 rather than ship a version with known bugs. Not wanting to
> get into the 1.7.x series is understandable, of course.

Believe me, it is not by choice.  

We were being held back by some architectures failing, or some dependent
libraries failing, and then most recently another bug in lib pulled in by
r-gnome. I spend a decent chunk of time on that in the last few weeks, and
with help from others (Chris Steigies and Kurt deserve special thanks for
their help with m68k where we switched back to f2c) are now in a position
where R 1.7.1 should appear in testing after the usual ten day delay, unless
bugs are found.  At least all architectures built within 24 hrs of my upload
which isn't bad.

[ In previous times we had other problems with the toolchain migration which
led to the separate testing versions Doug built for CRAN, with luck that
won't be required for a little while as those problems are now sorted out
and gcc 3.2/3.3 work uniformly across all our architectures, and required
libraries have caught up. ]

Dirk

-- 
Don't drink and derive. Alcohol and analysis don't mix.